According to these fonts' properties:

Cambria: "Authors: Monotype Imaging and Tiro Typeworks".
Calibri: "Authors: Luc(as) de Groot"

Theunis — Cambria was designed by Jelle Bosma (with Steve Matteson & Robin Nicholas of Monotype). Not sure what Tiro is doing in the font properties. Are you sure you didn’t mistakenly check Constantia (John Hudson’s contribution) instead?

I’m always mixing up those C faces, myself.

Ross did most of the Cambria Math work, although some of it based on Jelle's initial work. I did most of the the languages extensions for the Win7 versions of Cambria Regular and Italic, with David Březina working on the bold weights.

Can you give examples, illustrations of the issue? Also, which versions of the fonts are you using? The initial, Windows Vista versions did not support mark positioning; the Win7 versions do.

Which Cambria version is the most recent? 5.93?

Thank you very much! I have to ask: when this gets fixed, will it be available as an Windows (Office) Update? How does Windows update its fonts? And also will this enable these alternates also to be used with the simple Serbian (Cyrillic)?

The issue in Calibri is similar. I will post a screenshot.

Could you answer my question: do you know how will Microsoft release the fix? Thank you.

I'll ask Microsoft to look into this Calibri issue. It looks like the substitution of the cap accent forms (ccmp) is working fine, but the positioning (mark) is not. Possibly this is another situation with a lookup not being correctly associated with the SRB language system.

This is not expected to work in Constantia: of the original ClearType font collection, only Cambria, Calibri and Consolas were extended and mark positioning added for Windows 7.

I don't know how Microsoft will release a fix. Typically, critical bugs get rolled out using the Windows Update mechanism, but font issues are seldom considered critical in this sense.

Oh, OK. Thank you, John! I figured that it will ship with the next installment of Windows, as Sii says.

As for Constantia, I was just wondering why do the combining diacritical marks exists when they don't work properly. I thought typeface designers checked if all of that worked well. Segoe is another such typeface.

Oh, I see. Of course. Thank you for all your help and explanations!

Here is what I get with XeLaTeX (Russian left, Serbian right) using Cambria italic.

\documentclass[12pt]{article}
\usepackage{polyglossia} 
\setmainfont[Script=Cyrillic]{Cambria}
\setdefaultlanguage{russian}
\setotherlanguages{serbian}
\newfontfamily\serbian[Script=Cyrillic,Language=Serbian]{Cambria}
\begin{document}%\color{blue}
\large\itshape
м̥р̥д̥б̥\quad\textserbian{м̥р̥д̥б̥}
\end{document}

That is the full input.

Sylph: Also, in Cambria, is the Combining Ring Below supposed to look like this?

What is happening here is that the GPOS mark positioning is moving the mark beyond the OS/2 table usWinDescent value. This means that the mark will be clipped in some software, but not in others. If you are using MS Word, you can prevent this from happening by using an exact size of linespacing in the paragraph settings, rather than a multiple of the default linespacing.

In Michel's image, I'd say that the position of the mark under the default italic д is a bug. Probably my fault, I'm afraid. I think the mark height must be inherited from the upright font, where it needs to be lower because of the descenders of the д.

I have to admit that I've almost given up trying to understand how MS Word determines when and where to clip. Setting the linespacing to an exact value appears to do a better job avoiding clipping of marks above than below. Testing the sequence you show, less of the ring is clipped as the linespacing gets larger, but obviously this isn't a viable solution.

Note that this clipping should not occur when you print the document, including printing to PDF: it is the result of how Word paints each line of text on screen.

I'm not able to reproduce the problem you show with Calibri à vs а̀: these display identically in Word 2007 on my system.

Look closely at your image, above: the grave accent on the right appears to be from a different font, so maybe what you are seeing is a copy/paste error. Select the whole word and ensure that the font is Calibri.

I have to admit that I've almost given up trying to understand how MS Word determines when and where to clip.

But maybe some clever dude(s) or dudette(s) will remedy the situation with a pan-linguistic cross-platform script to help typographers by making that metadata available on the fly...

There is not need to even try understanding it. :) This is clearly an application bug and needs to be fixed by the application developers.
GPOS, and its arbitrary mark stacking mechanism in particular, by design (designed by Microsoft!) contradicts the idea that any font metadata would delimit the vertical region up (or down) to which attachments are visible. And likewise, the suggestion that OS/2 ascent and descent values would need to reflect every possibly additional mark is in contradiction to the idea and design of GPOS.

I have just checked the typeface version in Windows 8 Consumer Preview, and for Cambria it is still 5.96. Oh, Windows, will you update these fonts?

OK, the Cambria issues seem to be fixed in Windows 8 (albeit, one still has to pick a language for a non-existent state, you cannot get the right italics for Serbian (Cyrillic)), the Calibri misplacement of combining diacritics isn't fixed.

It seems also that both still don't have the U+203F character.

Not quite right, Herb. MS didn't replace the swastika with a star of David; in 2004 they removed two swastikas and the star of David from the font, presumably over GeoPol sensitivities.

[For another client, I once had to replace a six-pointed asterisk with a five-pointed on the grounds that the former might offend Arab customers by resembling the star of David. That was when I instituted the stupidity premium of +25%.]